The system, which is based on machine learning to analyze EEG signals, can predict epileptic seizures about an hour before they occur * NeuroHelp, the recently established start-up company that received a license to develop and commercialize the technology, is a portfolio company of Maitz Oasis, of a center Entrepreneurship 360 of Ben-Gurion University
Researchers at Ben-Gurion University of the Negev have developed an innovative and first-of-its-kind device for detecting and predicting epileptic seizures using machine learning algorithms. The wearable device can generate an early warning of an impending attack that will be sent to the cell phone about an hour before the attack begins. A license to develop and commercialize the system was given to NeuroHelp, a company recently established by me BGN Technologies, the technology commercialization company of Ben-Gurion University in collaboration with Dr. Oren Shariki from the Department of Cognitive and Brain Sciences at Ben-Gurion University and the scientific founder of NeuroHelp.
Epilepsy is a neurological disease with a widespread effect and varying degrees of severity. About 30% of patients do not respond optimally to drug treatment and live in constant fear of seizures. For these patients, an effective device for predicting seizures can significantly improve their quality of life and allow them to avoid injuries due to seizures.
The devices that exist today that warn of seizures can detect a seizure in real time but are unable to provide early warning of an impending seizure. The innovative device for predicting and detecting seizures is based on a unique combination of brain activity monitoring with the help of EEG and machine learning algorithms.
Dr. Oren Shariki said, "Epileptic seizures expose epileptic patients to a variety of preventable risks, including falls, burns and other injuries. Unfortunately, today there are still no devices for predicting seizures that can alert patients to an impending seizure and allow them to prepare for it. That is why we are excited that the machine learning algorithm we developed can accurately predict upcoming attacks, up to about an hour before they occur. Since we have also shown that our algorithm supports a significant reduction in the number of EEG electrodes required, the device we are developing will be both accurate and user-friendly. We are currently developing a prototype that will be ready for evaluation in human clinical trials later this year."
The system combines an EEG device with advanced software that minimizes the number of necessary EEG electrodes and provides information on the optimal placement of the electrodes on the scalp. The sophisticated algorithms filter the background noises that are not related to brain activity, extract informative measures of brain activity and distinguish between brain activity before a seizure and brain activity that does not precede a seizure.
Dr. Hadar Ron, chairman of the board of NeuroHelp, noted, "Epilepsy that does not respond optimally to medication characterizes up to 30% of epilepsy cases, and for that reason an accurate and easy-to-use device for predicting seizures is a real unmet medical need. The devices that exist today to warn of seizures detect the seizure at the time of its occurrence, and most of them depend on changes in movement, such as muscle spasms or falls. Our device is unique in that it can predict seizures and allow patients and their care team to take proactive actions to prevent injuries. It is also the only non-invasive device, which is based on brain activity and not on muscle movements or heart rate. We are confident that it will be a significant device in the management of epilepsy that is resistant to drug treatment."
Josh Peleg, CEOTechnologies BGN, the technology commercialization company of Ben-Gurion University, added, "NeuroHelp, a subsidiary of Technologies BGN, was recently founded as part of an accelerator Oasis of Ben-Gurion University, to further develop and commercialize their innovative solution for the benefit of people suffering from epilepsy. Just recently, NeuroHelp won the first investment award in SiliconGev's start-up companies competition, and this award is an important recognition of the extraordinary potential of this technology, which is based on a unique combination of brain research and artificial intelligence developed in Dr. Shariki's laboratory."
The new algorithm was developed and tested on EEG data from a large database of people with epilepsy whose brain activity was monitored for several days before surgery. The patients' data were divided into short segments of pre-seizure or during-seizure activity. Several machine learning algorithms with varying levels of complexity were trained on a database initially allocated for training (80% of the total EEG data in the pool), and then their ability to predict a seizure was tested on the remaining EEG data (20% of the total EEG data in the pool). Also, the effectiveness of the different algorithms was tested depending on the number of EEG electrodes. The algorithm with the best predictive ability reached an accuracy level of 97%, and its predictive ability was almost unimpaired (95% accuracy) even when it relied on a relatively small number of electrodes.
Epilepsy is a chronic non-infectious brain disease that affects approximately 65 million people worldwide. It is characterized by repeated seizures, which are short events of involuntary movements that may involve one area of the body or the entire body, often accompanied by loss of consciousness and loss of control of the sphincter. The seizures are the result of increased electrical activity of a group of brain cells. Different areas of the brain can be sites for this increased activity. Seizures can range from a brief loss of consciousness or a small muscle spasm to severe and prolonged spasms. The frequency of attacks also varies, ranging from less than one a year to several attacks per day.
More of the topic in Hayadan:
2 תגובות
Dogs do the job just as well as electronic devices can predict seizures
Better pacemakers or electrodes have not been proven to have a ton of associated problems
We don't always give the opinion about the quality of life of an epilepsy patient
The invention is fascinating and interesting. The problem is that I understand that you have to attach some electrodes to your head and that is what is discouraging. It is possible that the invention is suitable for patients who have several attacks a day and are anyway at home under supervision. If you manage to operate the device wirelessly then it will be an invention worthy of a Nobel Prize. Perhaps it is possible to invent a small device that will be aesthetic and will hang on the neck close to the wrong nerve and it will transmit electrical activity to your device and then the algorithms will interpret it as a pre-seizure state. Or combine your technology with the VNS technology. My wife told me she would not agree to stick electrodes on the head. The EEG test was a nightmare. We had a dog who was able to distinguish a seizure by smell and not by training. Is it possible to develop a device that also smells the changes in body odor and gives an additional indication before an attack? I would love to receive updates and be the first to purchase your device if you solve the electrode problem. Don't forget that most epilepsy patients try to deny the disease and then the electrodes actually with all the benefits also betray the disease at the same time. In any case, I have huge respect for the research itself. My wife is an epileptic and most of the time she is unbalanced on and off. There are also days of stability. That's why this topic is close to my heart. I'm racking my brain how to help her (checking food, sleeping hours, hormonal status, etc.) and at a time of instability the nervousness increases and goes when I'm not able to change the situation. I am neither a scientist nor a doctor. Thank you very much!